I ran this python script on https://dumps.wikimedia.org/wikidatawiki/entities/latest-all.json.bz2 and it ran for about 5 hours. ( did not complete but I terminated it, as the json dump in about 62G of data) . Nevertheless for those 5 hours I did not get a malformed precision in the output ( badprecisions.txt ) fille. in otherwords they are extremely rare in the database or they are not present or the script is missing something?.

import sys
import json
import os
import bz2
import gzip
GlOBE_COORDINATE_PROPERTIES= [
    'P625', 'P626', 'P1259', 'P1332',
    'P1333', 'P1334', 'P1335', 'P2786',
    'P5140', 'P8981', 'P9149'
]
def read_dump(path):
    mode = 'r'
    file_ = os.path.split(path)[-1]
    if file_.endswith('.gz'):
        f = gzip.open(path, mode)
    elif file_.endswith('.bz2'):
        f = bz2.BZ2File(path, mode)
    elif file_.endswith('.json'):
        f = open(path, mode)
    else:
        raise NotImplementedError(f'Reading file {file_} is not supported')
    try:
        for line in f:
            if isinstance(line, bytes):
                line = line.decode('utf-8')
            try:
                yield json.loads(line.strip().strip(','))
            except json.JSONDecodeError:
                continue
    finally:
        f.close()

with open('badprecisions.txt', 'w') as f:
    f.write('')


for item in read_dump(sys.argv[1]):
    precision = 0
    for geoProperty in  GlOBE_COORDINATE_PROPERTIES:
	    for claim in item.get('claims', {}).get(geoProperty, []):
	        try:
	            precision = claim['mainsnak']['datavalue']['value']['precision']
	        except:
	            continue
                 if precision >= 360 or precision <= -360:
	                with open('badprecisions.txt', 'a') as f:
	                    f.write(item['id']  '\t' str(precision)  '\n')

There was a report on Contact_the_development_team about the coordinates of St John's College (Q691283).

Maybe it's relevant these were added in June 2013.

It should at least find https://www.wikidata.org/wiki/Q3642430 too.

I got a list of Items that give a precision = null for statements with coordinate location(P625), But for some reason, Q3642430 is not part of the list.

https://phabricator.wikimedia.org/P16319

Three notes:

• Please edit your message and link to the paste instead of transcluding it (with {}). It's around 100K lines and It takes really long time to open this ticket now and my computer basically freezes because it's massive
• We need to check for more than P625, you can check for datatype in the snak instead of property
• The ones that have None are interesting but the main one is the ones that are not between -365 and 365. It seems the json output just gives an error: https://www.wikidata.org/wiki/Special:EntityData/Q3642430.json

new script https://gist.github.com/rosalieper/83354c6677c92a008f412b21d7d41575. Thanks for the feedbacks @Ladsgroup. I Will post the results of the script once it is done running.

There are 2 affected items with malformed coordinates.

Item ID--------- Precision

Q3629997---- 146706019195900
Q3642430---- 146706019195900

The script looks good and the findings are consistent with error messages found in logstash: 262 corresponding "RDF builder" error messages have been logged in the last 90 days, but none are found when excluding these two item IDs.

          #
  #      ##
  #     # #
#####     #
  #       #
  #       #
        #####

In T283576#7141092, @Ladsgroup wrote:

Three notes:

• Please edit your message and link to the paste instead of transcluding it (with {}). It's around 100K lines and It takes really long time to open this ticket now and my computer basically freezes because it's massive
• We need to check for more than P625, you can check for datatype in the snak instead of property
• The ones that have None are interesting but the main one is the ones that are not between -365 and 365. It seems the json output just gives an error: https://www.wikidata.org/wiki/Special:EntityData/Q3642430.json

Why -365 and 365? If the precision is in degrees (as suggested by Wikibase DataModel § Geographic locations, then I assume that range should use 360, not 365. But also… what’s a negative precision supposed to mean anyways? (I don’t really understand this field, I think.)